AI AND HUMANS: CAN THEY REACH A MARKETING BREAKTHROUGH TOGETHER?

Dmytro Bilash (co-founder Captain Growth) shared his professional opinion on what is the next marketing breakthrough and how will  humans and machines cooperate in the age of machine learning and artificial intelligence during his inspiring talk at GROWTH stage at EMERGE 2019.

Dmytro is an incremental entrepreneur and has vast experience in ad technology. His expertise lies in the intersection between marketing, data science, and entrepreneurship. Dmytro started his entrepreneurial career in a digital marketing agency. Later on, he decided to create his own startup. In 2017 he co-founded a company called Captain Growth, an AI tool with predictive analytics to help businesses to run their ads better.

The current digital marketing mix paradigm consists of search ads, email promotions, social media (Facebook, Instagram, Twitter, etc.), display ads and website content.

This is what all businesses have now. But we should always keep in mind that the marketing environment is changing rapidly. New players are coming to the field and they put new things on the table in front of marketers. Thus, in the future, we will have the following picture in digital marketing.

Messenger ads, bots, marketplace apps, new social media platforms, such as TikTok, Snapchat, they are all essential parts of marketing games. Today each marketer runs on average 3−5 channels at once. Very soon there will be 7−10 channels, and all of them should be managed all together, Dmytro argued.

Therefore, we have a big marketing challenge to move from single-channel to cross-channel world. The majority of marketing managers have single-channel mindset, meaning that looking and analysing various digital media as separate systems. However, having in mind that all digital platforms are interconnected and users consume data across all networks at once, marketers should look at all channels as one big system. Eventually, we need to make decisions based on the cross-channel data we have.

Even though cross-channel marketing sounds like something bold and creative, but we must understand its complexity. Analyzing data in one particular channel is a complicated task, but when it comes to the cross-channel world, the multiplicity of such connections increases exponentially. There is an opinion that only AI solutions can help to handle a huge amount of cross-channel data at once.

There are two types of AI available for use, and they are both fighting to draw our attention and marketing budget. The first type is White-box AI and the second one is Black-box AI.

The main difference between the two approaches in AI lies in explainability. In the black-box, humans cannot explain how the machine came up with that particular solution. Let’s say we have an input, which is customer behavior. The output is simple — yes or no, either this particular customer buy this product or not. With the black-box solutions, decisions will be made automatically — you give data and in return receive an output. You will never know how and why the machine came up with that output, as you have no idea what was the reason behind processing each particular recommendation. You only see the result and decide whether it is relevant to your work or not.

With the white-box AI, the situation is completely different because the output can be explained. When applying the white-box approach it is pretty clear why the machine made such a recommendation based on a particular prediction for that case. Also, it is quite easy to track the logic of the platform from input to output, as you can see the steps AI made to come up with that decision.

Indeed, AI technologies gained more popularity over the last years. People outsource the decision-making processes to machines. Hence, the question about white-box and black-box approaches in AI attracts more attention. So, the issue of trust in AI is among the most important ones. If AI made a mistake, then who would be responsible for that? Most of the big companies are stressing the point that transparency, building trust with machines and making the solutions more white-box is necessary. Nevertheless, marketers still feel insecure and uncomfortable with AI that uses their budget without strict control.

1. AI doesn’t know the context. As a marketer, you see the big picture of your business and you have all the data in your mind. You definitely know what type of audience resists to use your product, but there is hope that quite soon something will be changed in the market. Machines don’t know that and it’s a huge drawback, which might lead your business in the wrong direction. Thus, understanding the context of your business is something that the machine cannot perform right now.

2. AI is not able to adapt very fast and it is not very flexible. AI is usually tied to each particular task and only focused on that.

3. AI cannot explain the results to your CEO and defend them. When you work in a big company you should communicate your results in front of other people and make presentations to your bosses. Obviously, AI is unable to do it.

1. In the white-box approach, the machine is able to explain why this particular decision has been made. In this direction, you can understand the reason behind each specific recommendation and build trust with the AI-powered product.

2. The black-box solutions, such as neural networks, are able to be trained. But once you train them, you cannot control the process, in the end, you can only check the results. If outcomes are not sufficient enough, you might retrain them until you receive the desirable outputs. With the white-box approach, it is different. It gives you tiny opportunities to adapt results acceding to your needs within a particular task.

3. As a marketer, you have to see the full picture of your business in order to understand what is the aim of each particular channel, their features and differences. Only the white-box AI has such an ability to understand the context of your business as well as consume and proceed with a big amount of data.

1. As one guru in data science said, 'no data — no business'. Without data, you are not able to implement any AI. In this way, businesses should think about data collection stack first of all: how they are going to collect data, store it and work with it in the most efficient way. After that, you need to perform data normalization, which is a must-do procedure for cross-channel businesses. Data from Facebook, Google, Amazon, Snapchat — these are all different types of data, and you need to somehow normalize it and put in the same buckets. Suchwise, each company should think of their own solutions on how to make all these data equal.

2. Decision-making stack and automation. There are various AI products that might help you with that. Once you collect data and extract the required information from there, AI will help you to proceed and make the right decision.

3. Prepare people to use AI in a proper way and become Iron Mans. Basically, there two types of people in the team — those who do not trust AI at all and the other ones who trust them too much. Dmytro believed that the right way to use AI is to be Tony Stark from the Iron Man. The costume without Tony Stark is not so cool, and Tony Stark would never be an Iron Man without his costume. The same with AI, it is able to cooperate with humans. It helps us to analyze a huge amount of data, but without us, as marketers, it is inefficient to make the right decisions. Thus, it is really important to educate tech specialists in the proper way and explain to them the difference in the area of responsibility of humans and AI.

At the end of his talk at EMERGE Dmytro wished everyone a good AI adoption. There is a hope that we will live in a much better world in a few years with AI onboard.


EMERGE team hopes you enjoyed reading this article.
Stay tuned!